Build an Eclipse development environment for Perl, Python, and PHP

Use the Dynamic Languages Toolkit (DLTK) to create your own IDE


Before you start

About this tutorial

This tutorial shows how Eclipse's Dynamic Languages Toolkit (DLTK) makes it possible to build development tools for scripting languages. In particular, it explains how to implement syntax coloring, user preferences, and interpreter integration in a plug-in-based project.


This tutorial explains—one step at a time—how to build a DLTK-based development environment. The discussion presents DLTK by focusing on a practical plug-in project based on the Octave numerical computation language. The topics covered include:

  • Creating a plug-in project
  • Configuring the editor and Eclipse DLTK text tools
  • Adding classes to control syntax coloring in the text editor
  • Enabling user preferences
  • Integrating the script interpreter into the development environment
  • Adding a custom console to communicate between the workbench and the interpreter


This tutorial is written for Java™ developers familiar with Eclipse and interested in building tools for dynamic languages, such as PHP, Tcl, Ruby, and Python. It assumes that you have a basic understanding of plug-ins and Eclipse-based development tools in general.

System requirements

To build the example project in this tutorial, you need a computer with an Eclipse installation (the examples in this tutorial use version 3.7 or later) and a current JRE. You must install a recent version of the DLTK, which we describe in the section "Install the DLTK." (See Related topics for links to download Eclipse.)

Introducing DLTK and the DLTK editor

Eclipse offers many capabilities for building custom development tools, but they're not easy to use or understand. Many of the classes, such as TextEditor, require significant configuration to work properly. In contrast, the DLTK provides a prepackaged development environment that requires only minor tweaking to be operational. Building a development tool with the DLTK is like baking a cake from a mix: The output isn't completely original, but you get a maximum of quality and reliability with a minimum of effort.

A second important difference between DLTK and Eclipse involves the supported languages. Eclipse's build capabilities are based on compiled languages like Java, C, and C++. The DLTK is geared toward dynamic languages, which include scripting languages like PHP, Ruby, and Tcl. Ideally, a development tool for these languages supports easy integration of script interpreters, command evaluation on a line-by-line basis, and the ability to enter commands at a console. Eclipse doesn't provide these features, but the DLTK supports them all.

Install the DLTK

If you haven't installed the DLTK, do so now. Perform the following steps:

  1. Open Eclipse, then click Help > Install New Software.
  2. Select Indigo - from the Work with list.
  3. Select the General Purpose Tools > Dynamic Languages Toolkit - Core Frameworks check box, then click Next.
  4. Click Next after reviewing the prerequisites that will also be installed.
  5. If you accept the license agreements, click Finish to install the plug-in.
  6. After installing the files, restart Eclipse.

The DLTK-based plug-in project: The Octave IDE

This tutorial does more than just discuss the DLTK and its features. It walks you through the steps of creating an DLTK-based IDE for the Octave scripting language. Octave is a GPL-licensed toolset for performing mathematical computation, particularly with matrices. Octave commands are easy to understand, and the language closely resembles that of the popular proprietary Matlab tool. You can download the example code from Download.

The Octave IDE you'll be building won't have all of Eclipse's bells and whistles, but you will be able to create Octave scripts (*.m files), edit them in a multifeatured text editor, run the Octave interpreter with the click of a button, and see the result displayed in the IDE's console. Figure 1 shows what the editor and console look like.

Figure 1. The Octave IDE
Screenshot of the Octave ID editing the file test.m with code in the upper window and console output in the lower window
Screenshot of the Octave ID editing the file test.m with code in the upper window and console output in the lower window

This tutorial explains the process of building the Octave IDE in four steps:

  1. Create the Octave editor and add features.
  2. Store and respond to user preferences.
  3. Integrate the Octave interpreter.
  4. Display the interpreter's output in the console.

The first task deals with the Octave script editor. There's a lot to learn about this subject, but before you start coding, let's take a step back and look at how DLTK editors work. Much of this discussion applies to all Eclipse text editors.

A high-level introduction to the DLTK text editor

At its most basic level, the DLTK text editor is just a regular SWT StyledText widget with a lot of additional capabilities. These capabilities fall into one of two categories:

  • Responding to user events. Changing the editor's display according to the user's keystrokes
  • File I/O. Transferring character data between the IDE and files on the user's system

In DLTK editors, event handling is managed by a ScriptSourceViewer object. This object, created by the ScriptEditor, wraps around the StyledText widget and manages its response to user keystrokes. Typically, this response involves modifying the widget's presentation or updating the editor's internal data. This data is embodied in an IDocument object, which not only stores text but also information related to line numbers, positions, and regions, called partitions.

The ScriptSourceViewer updates the IDocument by accessing an object called a SourceModuleDocumentProvider. In addition to providing access to the IDocument, this object handles file operations. DLTK handles all the editor's file interactions for you, but if you're interested in learning more, read the documentation on the object's superclass: TextFileDocumentProvider.

If you're familiar with the MVC pattern, the relationship between the ScriptSourceViewer, IDocument, and StyledText widget should be clear. The IDocument serves as the editor's Model aspect, containing data independent of its presentation. The StyledText widget represents the editor's View aspect, and the ScriptSourceViewer serves as the Controller, managing communication among the Model, the View, and the user. Figure 2 shows this relationship.

Figure 2. Primary elements of the DLTK editor
Diagram shows the IDocument (model) feeding into the ScriptSourceViewer (controller) which feeds the StyledText (view)
Diagram shows the IDocument (model) feeding into the ScriptSourceViewer (controller) which feeds the StyledText (view)

If the explanation thus far makes sense and Figure 2 seems reasonable, you should have no difficulty understanding the technical discussion that follows. In the next section, you start creating the Octave plug-in.

Getting started with a DLTK project

For the sake of simplicity, this tutorial integrates all the Octave IDE functionality into a single plug-in. Generally, you implement a feature using multiple plug-ins: one containing core classes, one containing UI classes, one containing debug classes, and so on. But this tutorial's code only needs one plug-in, so if you have Eclipse running, click File > New > Project, select the Plug-in Project option, and click Next to begin creating the plug-in project. Follow the steps here to create a plug-in project that includes a sample editor, which is a good starting point for your new plug-in.

  1. Enter the project name, which in this example is org.dworks.octaveide, as shown in Figure 3.
    Figure 3. Entering the project's name
    Screenshot shows the New Plug-in Project screen where the Project name is being entered as org.dworks.octaveide
    Screenshot shows the New Plug-in Project screen where the Project name is being entered as org.dworks.octaveide
  2. In the Content screen, shown in Figure 4, enter org.dworks.ocativeide as the ID if it's not already set, and click Next.
    Figure 4. The Content screen
    Screenshot shows the Content screen of the New Plug-in Project with the ID set to org.dworks.octaveide
    Screenshot shows the Content screen of the New Plug-in Project with the ID set to org.dworks.octaveide
  3. Select the Create a plug-in using one of the templates check box, select Plug-in with an editor from the list, and click Next.
  4. Enter OctiveEditor, as shown in Figure 5, enter m for the File Extension, and then click Finish to create your plug-in project.
    Figure 5. Entering the information for the editor
    Screenshot shows the Sample XML Editor screen with the File Extenion set to 'm'
    Screenshot shows the Sample XML Editor screen with the File Extenion set to 'm'

Your new plug-in project is created with the plugin.xml configuration file already available in your project, and you're off to a good start. You can edit the plugin.xml configuration using the built-in editor, or you can edit the source of the file directly by clicking the editor's plugin.xml tab. This tutorial shows you the listing of the plugin.xml file itself.

Before continuing with the rest of the tutorial, you should add two dependencies that allow your plug-in code to find the DLTK classes. In the Dependencies tab of your plugin.xml editor, click Add, and add org.eclipse.dltk.core and org.eclipse.dltk.ui. The current version in this tutorial is 3.0.1.

Configure the Octave plug-in

The plugin.xml file that was started when you created your plug-in project contains information about the plug-in's extensions. These extensions tell Eclipse how the plug-in intends to contribute to the workbench. To start, the plugin.xml file needs two simple extensions: one that identifies the Octave editor and one that identifies the types of files to be edited. The first extension was already created when you created the plug-in project using the template.

Listing 1 presents the first extension, which defines characteristics of the Octave editor.

Listing 1. The Octave IDE editor extension
<extension point="org.eclipse.ui.editors">

This extension contains several important attributes. The class attribute defines the primary class that embodies the editor: OctaveEditor, which you will create shortly. The contributorClass attribute identifies the class that provides actions for the editor. And the default attribute states that the OctaveEditor should be the default editor for associated files. The name attribute identifies the text that will be displayed for the editor. As defined in the file, the property corresponds to "Octave Script Editor."

The last supplement in the extension, contentTypeBinding, requires explanation. Eclipse maintains a registry of content types that represent file formats and naming conventions. By adding a contentTypeBinding element to the editor's extension, you can associate the editor with file suffixes, file names, and even file aliases. But for your Octave editor, you're only interested in *.m files, so your extension of org.eclipse.core.runtime.contentTypes is simple. Listing 2 shows this second extension.

Listing 2. The Octave IDE ContentType extension
<extension point="org.eclipse.core.runtime.contentTypes">

Many more attributes are available for customizing content types, but this declaration identifies the org.dworks.octaveide.octaveContentType content type as being text based and associated with the file suffix .m. The priority attribute states that if a file becomes associated with multiple content types, this type should be given high priority.

Creating a DLTK language toolkit class

A principal advantage of using the DLTK is that it's simple to customize how IDEs work. This customization is made possible through DLTK's language toolkit interfaces (IDLTKCoreLanguageToolkit, IDLTKDebugUILanguageToolkit, IDLTKUILanguageToolkit, and so on). These interfaces present methods that, when implemented, configure many aspects of the IDE, such as parsing, color usage, user preferences, and accessing the interpreter.

This tutorial focuses primarily on the DLTK UI, so you'll only concern yourself with the IDLTKUILanguageToolkit. The methods in this interface perform one of two functions: They either return a class that handles editor configuration or provide the ID of a configuration object declared in plugin.xml. Seven of these methods are listed here in Table 1.

Table 1. Methods on the IDLTKUILanguageToolkit interface
Method nameDescription
getEditorId(Object elementID)Returns the editor's ID
getTextTools()Returns the ScriptTextTools object that will customize how text is presented in the editor
createSourceViewerConfiguration()Returns the ScriptSourceViewerConfiguration object that will configure the operation of the editor's source viewer
getPreferenceStore()Returns the IPreferenceStore object that contains user settings
getEditorPreferencePages()Returns the IDs of the preference pages related to editor settings
getInterpreterPreferencePage()Returns the IDs of the preference pages related to the script interpreter
getIntepreterId()Returns the ID of the script interpreter

In the example code, the OctaveUILanguageToolkit class fleshes out these methods with objects and identifiers. The first two objects discussed relate to the first two methods in the toolkit: OctaveEditor and OctaveTextTools.

Create the Octave editor and the Octave text tools

Now that you've configured the editor's extension, the first order of business is to create the OctaveEditor. Thanks to the DLTK, it won't require much coding: The DLTK's ScriptEditor class (the superclass of OctaveEditor) handles most of the editor's operations on your behalf. All you need are a few configuration methods. In the OctaveEditor, the most important of these is getTextTools(), which returns a ScriptTextTools object.

The ScriptTextTools class is convenient. This central class accesses several other classes that provide the editor's capabilities. By subclassing ScriptTextTools, an editor can customize how these capabilities are performed. The Octave editor creates a subclass called OctaveTextTools, which provides access to the important objects shown in Table 2.

Table 2. Objects made available by OctaveTextTools
OctavePartitionScannerReads the editor text and determines partition boundaries
OctaveSourceViewerConfigurationConfigures the ScriptSourceViewer and the way it responds to user-generated events

These objects become particularly important in implementing syntax coloring—that is, changing the color of text depending on the text's syntactic function in code. The next topic explains this feature in greater detail.

Syntax coloring in a DLTK editor

Syntax coloring is one of the most noticeable aspects of any professional source code editor. Not only do the colors make the code easier to read, but they also make typing errors readily apparent. Figure 6 shows how comments, numbers, keywords, and strings are colored differently from regular code.

Figure 6. Syntax coloring in the Octave IDE
Screenshot shows code in the Octave IDE with comments in green, keywords in dark blue, string in cyan and values in red
Screenshot shows code in the Octave IDE with comments in green, keywords in dark blue, string in cyan and values in red

Configure syntax coloring in the Octave editor

The DLTK process of syntax coloring consists of four steps:

  1. The ScriptSourceViewer responds to keystrokes by alerting the document's partitioner.
  2. The partitioner splits the document's text into partitions according to rules.
  3. When the partitioning is finished, the viewer alerts a presentation reconciler to analyze the modified partition.
  4. The reconciler calls upon a damager/repairer to analyze the affected region and apply syntax coloring.

In this manner, the editor uses a divide-and-conquer strategy to color text. First, it splits the document into partitions. Then, it analyzes the modified partition and determines whether and where syntax coloring should be updated. The next two sections explain this process in greater depth and show how to implement them in the Octave IDE.

Document partitioning

In addition to text, an IDocument stores an array of Positions that serve as boundaries of regions within text. These regions not only make syntax coloring possible (comments one color, code another color) but also let you apply different tools to different parts of the text. For example, there's no reason to tell your script parser to analyze commented lines. Using partitions, you can make sure your parser only operates on regions containing actual code.

Document partitioning is accomplished by two central objects: one that implements IDocumentPartitioner and one that implements IPartitionTokenScanner. The partitioner delivers text to the scanner, which analyzes the text and produces ITokens. The partitioner uses these tokens to set Positions within the document.

It would be nice if the OctaveEditor automatically initialized the partitioner and scanner whenever the user opens an *.m file. But the document's configuration must be explicitly defined in plugin.xml with an extension of the org.eclipse.core.filebuffers.documentSetup extension point. This extension identifies the class that should be called on to configure the document during its initialization. Listing 3 presents the extension defined for the Octave IDE plug-in.

Listing 3. Configuring the document
<extension point="org.eclipse.core.filebuffers.documentSetup">

The OctaveDocumentSetup calls on the DLTK ScriptTextTools object to initialize *.m documents. By default, this creates a FastPartitioner to serve as the document's partitioner.

Each DLTK editor has to create its own IPartitionTokenScanner object to read characters and produce tokens corresponding to partitions. For example, if you want to create a partition for text starting with /* and ending with */, you need to configure the scanner so that it forms the appropriate token when it encounters this text pattern. Table 3 lists how the Octave plug-in divides text into three partitions.

Table 3. Partitions into which the Octave plug-in divides text
CommentsIdentified by OCTAVE_COMMENT, which equals "__octave_comment__"
StringsIdentified by OCTAVE_STRING, which equals "__octave_string__"
Regular script codeIdentified by IDocument.DEFAULT_CONTENT_TYPE, which equals "__dftl_partition_content_type"

These partitions are defined in the IOctavePartititons interface. They are used to initialize the FastPartitioner, which sends them on to the partition scanner.

Partitions in the Octave editor can be identified using a set of logical rules, so your OctavePartitionScanner will extend the RuleBasedPartitionScanner class. The most important method in this class is setPartitionRules, which accepts an array of IPredicateRule objects. The goal of a rule is to convert sections of text into ITokens. Rules can be difficult to code from scratch, but thankfully, Eclipse provides many useful implementations of the IPredicateRule interface. The Octave IDE uses two of them to recognize partitions: SingleLineRule and MultiLineRule. Both classes have similar constructors, and they are initialized with the following information (at least):

  • The character that begins the partition
  • The character that ends the partition
  • The IToken corresponding to the recognized partition

For example, say you want to create a rule that produces a string token when it finds text starting and ending with a single quotation mark ('). You do this with the following code:

IToken stringToken = new Token(OctavePartitions.OCTAVE_STRING);
new SingleLineRule("'", "'", stringToken, '\\');

Listing 4 presents the entire code that OctavePartitionScanner uses to detect partitions in Octave scripts. The Octave editor creates comment partitions when it encounters text preceded by a percent sign (%) and followed by a new line. String partitions are delimited by single quotation marks.

Listing 4. The Octave IDE partition scanner
// Declare the different tokens
IToken string = new Token(IOctavePartitions.OCTAVE_STRING);
IToken comment = new Token(IOctavePartitions.OCTAVE_COMMENT);

// Create the list of rules that produce tokens
List<IPredicateRule> rules = new ArrayList<IPredicateRule>();
rules.add(new SingleLineRule("%", "\n", comment));
rules.add(new SingleLineRule("'", "'", string));
IPredicateRule[] result = new IPredicateRule[rules.size()];

Having created the OctavePartitionScanner and initialized its rules, you've finished configuring the editor's partitioning. The next step is to tell the editor how to scan and color text inside the partitions.

Scan text inside partitions

As mentioned, DLTK editors rely on a ScriptSourceViewer to respond to user events like keystrokes. You can tailor the viewer's operation with a ScriptSourceViewerConfiguration object. Like the ScriptTextTools class, the configuration object does little work of its own. Instead, its methods identify classes that assist the source viewer's operation. When it comes to syntax coloring, two of the most important of these "assistance" methods are shown here in Table 4.

Table 4. Assistance methods for syntax coloring
Method nameDescription
getPresentationReconciler()Returns an IPresentationReconciler that controls how text is presented when changes occur
initializeScanners()Returns an array of AbstractScriptScanners that perform the same type of text scanning as the RuleBasedPartitionScanners described earlier

These two methods are closely related. The following discussion explains how the Octave IDE implements them to provide text coloring.

The IPresentationReconciler

In Eclipse parlance, a reconciler is an object that keeps track of a document's content and responds to changes. A presentation reconciler is a reconciler that responds to document changes by altering how text is presented. Specifically, the IPresentationReconciler responds to changes by calling on the objects shown in Table 5.

Table 5. Objects used by IPresentationReconciler
Object nameDescription
IPresentationDamagerDetermines the extent of the document modification
IPresentationRepairerCreates a TextPresentation object to respond to the modification identified by the IPresentationDamager

These two objects operate in sequence. When the user modifies text in a partition, the IPresentationReconciler gives the partition name to the IPresentationDamager. After the damager identifies the modified region, the IPresentationRepairer analyzes the region to see if, where, and how the partition's text presentation should be updated.

To keep matters simple, Eclipse provides a DefaultDamagerRepairer that serves as damager and repairer. The Octave IDE constructs a DefaultDamagerRepairer for each of the three partitions and initializes each with a different scanner. This reconciler is shown in Listing 5.

Listing 5. Configuring the presentation reconciler
// Create a DefaultDamagerRepairer for the default partition (code)
DefaultDamagerRepairer dr = new DefaultDamagerRepairer(codeScanner);
reconciler.setDamager(dr, IDocument.DEFAULT_CONTENT_TYPE);
reconciler.setRepairer(dr, IDocument.DEFAULT_CONTENT_TYPE);

// Create a DefaultDamagerRepairer for string partitions
dr = new DefaultDamagerRepairer(stringScanner);
reconciler.setDamager(dr, OctavePartitions.OCTAVE_STRING);
reconciler.setRepairer(dr, OctavePartitions.OCTAVE_STRING);

// Create a DefaultDamagerRepairer for comment partitions
dr = new DefaultDamagerRepairer(commentScanner);
reconciler.setDamager(dr, OctavePartitions.OCTAVE_COMMENT);
reconciler.setRepairer(dr, OctavePartitions.OCTAVE_COMMENT);

Just as the partition scanner identifies partitions using rules, each of these scanners creates tokens according to similar rules. The next section discusses text scanners in detail.

Scanning text inside partitions

The constructor of the ScriptSourceViewerConfiguration calls initializeScanners() to create the scanners needed to analyze text. Listing 6 presents the full implementation of initializeScanners() in the OctaveSourceViewerConfiguration class.

Listing 6. Creating the script scanners
private AbstractScriptScanner codeScanner, stringScanner, commentScanner;

protected void initializeScanners() {

   // Create a scanner for the script code
   codeScanner = new OctaveCodeScanner(getColorManager(), fPreferenceStore);
   // Create a scanner for string partitions
   stringScanner = new SingleTokenScriptScanner(getColorManager(), 
fPreferenceStore, IOctavePartitions.OCTAVE_STRING);
   // Create a scanner for comment partitions
   commentScanner = new SingleTokenScriptScanner(getColorManager(), 
fPreferenceStore, IOctavePartitions.OCTAVE_COMMENT);

Because the partition scanner already identified strings and comments, these partitions don't need to be reanalyzed. This is why the stringScanner and commentScanner are both SingleTokenScriptScanners. This class doesn't make decisions based on rules but always returns the same type of IToken. For example, the stringScanner returns a token initialized with OctavePartitions.OCTAVE_STRING, and the commentScanner returns a token initialized with OctavePartitions.OCTAVE_COMMENT. You'll see how these token properties determine text color shortly.

The codeScanner is more involved and performs the brunt of the IDE's text analysis. This scanner is an OctaveCodeScanner and, like the PartitionTokenScanner, creates tokens based on rules. But instead of the MultiLineRule or SingleLineRule, it relies on the IRule implementations shown in Table 6.

Table 6. IRule implementations used by OctaveCodeScanner
WordRuleReturns a token when the scanner encounters specific words
NumberRuleReturns a token when the scanner encounters numbers

These rules are created in the createRules method of the OctaveCodeScanner. Listing 7 presents this method in full.

Listing 7. Creating the script scanners
protected List<IRule> createRules() {

   // Create tokens
   IToken keywordToken = getToken(DLTKColorConstants.DLTK_KEYWORD);
   IToken numberToken = getToken(DLTKColorConstants.DLTK_NUMBER);
   IToken defaultToken = getToken(DLTKColorConstants.DLTK_DEFAULT);

   // Create and populate list
   List<IRule> ruleList = new ArrayList<IRule>();

   // Create word rule to detect keywords
   WordRule wordRule = new WordRule(new OctaveWordDetector(), defaultToken);
   for (int i = 0; i < IOctaveKeywords.keywords.length; i++)
      wordRule.addWord(IOctaveKeywords.keywords[i], keywordToken);

   // Create number rule to detect numbers
   NumberRule numberRule = new NumberRule(numberToken);

   // Set the token returned for default text
   return ruleList;

It's important to distinguish the tokens created in Listing 7 from those created in Listing 4. The Token constructor accepts any Object as its argument, and this Object tells the token's recipient what action to take. In Listing 4, Tokens are initialized with partition names that tell the FastPartitioner how to update the document's partitioning. In Listing 7, Tokens are initialized with TextPresentation objects, which tell the IDE how to present text in the editor.

Listing 7 calls getToken to associate each Token with a TextPresentation. This method searches through a Map populated with the user's preferences. To understand how this map is updated, it's important to understand how DLTK preferences work. This is the subject of the following discussion.

Configuring DLTK preferences

The previous discussion showed how text presentation depends on the user's preferences. Now, we will go farther, showing how to create pages that receive the user's preferences. You can access these preferences by clicking Window > Preferences in Eclipse. At the conclusion of this discussion, we explain how the user's preferences make it possible to integrate the script interpreter into the IDE.

Preference stores

An Eclipse preference store is like a property file: It contains name-value pairs that identify the user's preferred settings. But there are two important differences. Each preference holds a default value in addition to its current value, and preference values must take one of six basic types:

  • Boolean
  • int
  • long
  • float
  • double
  • String

Preferences are stored and retrieved by calling methods defined by the IPreferenceStore interface. The setValue and setDefaultValue methods add and update preferences. To retrieve a preference value, you can call one of getBoolean, getInt, getLong, getFloat, getDouble, or getString. Similar methods access default values, such as getDefaultDouble.

To access preferences related to SWT objects (more particularly, RGB, Rectangle, FontData, and FontData[] objects), you can call methods in the PreferenceConverter class.

During initialization, a DLTK ScriptEditor creates an ArrayList of four IPreferenceStore objects:

  • The ScopedPreferenceStore created for the Octave plug-in (org.dworks.octaveide)
  • The ScopedPreferenceStore created for the Eclipse text editor plug-in (org.eclipse.ui.editors.text)
  • An EclipsePreferencesAdapter that provides preferences related to the project containing the script being edited
  • A PreferencesAdapter that contains the preferences associated with the DLTK core

These stores are combined into a ChainedPreferenceStore. If two or more preference stores contain identical preferences, the value is supplied by the store that added first.

Initialize preferences

Eclipse makes it possible to set preferences during initialization with the extension point org.eclipse.core.runtime.preferences. The Octave IDE extends this point and identifies the OctavePreferenceInitializer class as providing default values for the IDE's preferences, which in turn extends the abstract class AbstractPreferenceInitializer, whose only required method is initializeDefaultPreferences. The IDE invokes this method during initialization, and Listing 8 shows how it's implemented in code.

Listing 8. Initializing default preferences
public void initializeDefaultPreferences() {
   IPreferenceStore store = OctavePlugin.getDefault().getPreferenceStore();

This simple routine accesses the Octave plug-in's preference store and calls on the OctavePreferenceConstants class to initialize the store's values. This is a subclass of DLTK's PreferenceConstants class, which defines and initializes a wide range of general scripting preferences. In this tutorial, the only preferences that will be initialized are those dealing with syntax coloring. Listing 9 shows how this initialization is accomplished.

Listing 9. Setting syntax-coloring preferences
public static void initializeDefaultValues(IPreferenceStore store) {

   // Set default preferences for the editor

   // Make keywords blue and bold
   PreferenceConverter.setDefault(store, DLTKColorConstants.DLTK_KEYWORD, 
      new RGB(40, 0, 200));
   store.setDefault(DLTKColorConstants.DLTK_KEYWORD + 
      PreferenceConstants.EDITOR_BOLD_SUFFIX, true);

   // Set default values for other preferences
      DLTKColorConstants.DLTK_SINGLE_LINE_COMMENT, new RGB(25, 200, 25));
   PreferenceConverter.setDefault(store, DLTKColorConstants.DLTK_NUMBER, 
      new RGB(255, 25, 25));
   PreferenceConverter.setDefault(store, DLTKColorConstants.DLTK_STRING, 
      new RGB(50, 100, 100));

The first of the Octave-specific preferences deals with keywords. The code configures both the color and the style (bold, italics, strikethrough) of keywords in the IDE and requires two steps. First, the PreferenceConverter is called on to associate DLTKColorConstants.DLTK_KEYWORD with blue. Next, the store.setDefault method associates DLTKColorConstants.DLTK_KEYWORD + PreferenceConstants.EDITOR_BOLD_SUFFIX with true. These two lines ensure that if any Token is initialized with DLTKColorConstants.DLTK_KEYWORD, the corresponding text is displayed in blue and boldface.

Preference pages and configuration blocks

Now that you understand how to initialize preferences, it's important to give users the opportunity to configure preferences themselves. Eclipse makes this possible through preference pages. DLTK simplifies the process of constructing these pages with objects called configuration blocks. This section describes both of these classes and how they work together to provide user preferences in the Octave IDE.

Eclipse preference pages

When you click Window > Preferences in Eclipse, the left pane of the Preferences window presents a hierarchy of user-configurable options. For the Octave IDE, you want a top-level preference called Octave, with sub-preferences for syntax coloring and the Octave interpreter. Figure 7 shows what this window should look like.

Figure 7. The Octave Preferences window
Screenshot of Octave Preferences with tree of options on the left and the options for Syntax coloring, which is selected, on the right
Screenshot of Octave Preferences with tree of options on the left and the options for Syntax coloring, which is selected, on the right

In the Octave project, plugin.xml defines three preference topics by extending the point org.eclipse.ui.preferencePages. Each extension requires id, name, and class attributes. The two sub-preferences have an additional attribute—category—that names the id of the parent preference. These extensions are shown in Listing 10.

Listing 10. Octave preference extensions
<extension point="org.eclipse.ui.preferencePages">

The class attribute names a class that implements IWorkbenchPreferencePage, a subinterface of IPreferencePage. DLTK provides its own implementation of IWorkbenchPreferencePage called AbstractConfigurationBlockPreferencePage. DLTK also provides several subclasses for specific purposes. These class-interface relationships are depicted in Figure 8.

Figure 8. Preference page class hierarchy
Chart shows how the AbstractScriptDebug, AbstractScript, ScriptCore and ToDoTaskAbstract preference pages feed the AbstractConfigurationBlock preference page, which feeds the PreferencePage and IWorkbenchPreferencePage interface which both feed into the IPreferencePage interface
Chart shows how the AbstractScriptDebug, AbstractScript, ScriptCore and ToDoTaskAbstract preference pages feed the AbstractConfigurationBlock preference page, which feeds the PreferencePage and IWorkbenchPreferencePage interface which both feed into the IPreferencePage interface

The three classes in Listing 10 are all direct subclasses of AbstractConfigurationBlockPreferencePage. If you look through their Java files, you'll see they don't contain much code. Each class provides an ID, a label, and access to the plug-in's preference store. Further, every AbstractConfigurationBlockPreferencePage must implement the method createConfigurationBlock(), which constructs the object that provides the page's graphical aspect. This object must implement the DLTK interface IPreferenceConfigurationBlock. This important interface is discussed next.

DLTK configuration blocks

When it comes to presenting IDE preferences, the hardest part is building the UI. For example, the Octave syntax-coloring preference page not only allows the user to choose text colors and styles, it also provides an embedded text editor that shows how the styling actually looks. Creating these graphical controls from scratch can be time-consuming, but DLTK's IPreferenceConfigurationBlocks make your life much easier. Figure 9 shows what the configuration block class hierarchy looks like.

Figure 9. Configuration block class hierarchy
Image shows configuration block class hierarchy
Image shows configuration block class hierarchy

The syntax-coloring configuration block

The first class of interest is AbstractScriptEditorColoringConfigurationBlock, which was created specifically for providing the user with syntax-coloring preferences. The createSyntaxPage method performs the main work, which is to create the graphical preference page. More specifically, it forms a TreeViewer that displays the different syntax types and a set of buttons that represent color and style preferences. Then, it calls createPreviewer to build the embedded text editor.

The Octave plug-in code contains a subclass of AbstractScriptEditorColoringConfigurationBlock called OctaveSyntaxColorConfigurationBlock. Besides the constructor, the methods listed in Table 7 are implemented.

Table 7. The methods implemented by OctaveSyntaxColorConfigurationBlock
Method nameDescription
createPreviewViewerReturns a viewer to wrap around the embedded text editor
createSimpleSourceViewerConfigurationReturns a basic object to configure the viewer in the embedded text editor
setDocumentPartitioningCalls on an IDocumentSetupParticipant to set up the partitioning of the document
getSyntaxColorListModelReturns a 2D array that identifies the different types of syntax that can be colored and the corresponding key in the preference store

The first three methods are straightforward, because the embedded text editor doesn't need a complicated viewer, viewer configuration object, or document setup participant. However, the last method, getSyntaxColorListModel, requires explanation. This method provides a 2D array whose elements correspond to the syntax types in the page's TreeViewer. Each element requires three pieces of information: a name, a category, and a corresponding key in the page's preference store.

Going back to the example, Figure 10 shows that the Octave syntax color model consists of four different syntax types: keywords, numbers, strings, and single-line comments. The first three types fall under the General heading, while the last type falls under the Comment heading. The code in Listing 11 shows how these types are configured in OctaveSyntaxColorConfigurationBlock.getSyntaxColorListModel.

Listing 11. Creating the syntax color list model
protected String[][] getSyntaxColorListModel() {
   return new String[][] {	
      { "Single-Line Comments", DLTKColorConstants.DLTK_SINGLE_LINE_COMMENT, 
         sCommentsCategory },	
      { "Keywords", DLTKColorConstants.DLTK_KEYWORD, sCoreCategory },
      { "Strings", DLTKColorConstants.DLTK_STRING, sCoreCategory },
      { "Numbers", DLTKColorConstants.DLTK_NUMBER, sCoreCategory }};

When the user changes a preference setting in the page and clicks OK, the page updates the corresponding value in the preference store. The editor responds to the preference change by updating the presentation of its text.

Interpreter preferences

The last preference option in Figure 10 provides settings for the IDE's script interpreter. DLTK provides two classes that make this possible: ScriptInterpreterPreferencePage and InterpretersBlock. The InterpretersBlock creates the page's graphical control, which consists of a Label that provides instructions; a CheckboxTableViewer that lists available interpreters; and a series of buttons that allow the user to add, edit, copy, remove, or search for interpreters. Figure 10 shows what this page looks like along with the window that appears when the user clicks Add.

Figure 10. The interpreter preference page and window
Screenshot of the Add interpreter window with Interpreter type set to Octave Install
Screenshot of the Add interpreter window with Interpreter type set to Octave Install

The abstract class InterpretersBlock requires that subclasses implement two methods: getCurrentNature and createInterpreterDialog. The first method returns a String that represents the interpreter, the source code language, and other associated objects. This nature is frequently identified in DLTK extension points, and it's important. If your Eclipse installation contains multiple DLTK-based features for different dynamic languages, you can easily distinguish them by their natures.

The second method, createInterpreterDialog, forms the window that appears when the user clicks Add on the preference page. This window, shown in Figure 10, must furnish at least four pieces of information: the location of the interpreter executable, its name, its type, and the location of any libraries it requires to execute. To form this window, the createInterpreterDialog method requires an instance of the DLTK class AddScriptInterpreterDialog.

The AddScriptInterpreterDialog constructor accepts an array of interpreter types and an object that implements IInterpreterInstall. Having multiple interpreter types allows the IDE to support multiple interpreters for the same language. The Octave IDE only supports one installation type, and it's identified in plugin.xml with an extension of org.eclipse.dltk.launching.interpreterInstallTypes. These preference extensions are shown in Listing 12.

Listing 12. Octave preference extensions
<extension point="org.eclipse.dltk.launching.interpreterInstallTypes">

The class attribute identifies an object that implements IInterpreterInstallType. Each installation type is given a name, an ID, a nature (a String that identifies the interpreter and source type), and the locations of the interpreter and any required libraries. In addition, each IInterpreterInstallType must create one or more IInterpreterInstall objects.

Although an IInterpreterInstallType object defines a class of interpreter installations, an IInterpreterInstall object must be created for each particular installation. This object serves as the interpreter's primary data storage object, and it holds the information in Table 8.

Table 8. The data stored by IInterpreterInstall
NameA label for the interpreter installation
IDA logical identifier for the interpreter installation
NatureA logical identifier for the IDE (interpreter, source files, and so on)
Installation locationThe path to the installation of the interpreter
Library locationsPaths to the interpreter's required libraries
Execution environmentEnvironment variables and process configuration
Interpreter argumentsParameters directed to the interpreter

Finally, each IInterpreterInstall must provide access to an IInterpreterRunner object. This object is responsible for actually starting the interpreter, and the following discussion explains not only how it works but how it fits into the interpreter-IDE interaction as a whole.

The DLTK interpreter and console

The most important feature of a DLTK application isn't the text editor or preferences but the ability to launch an interpreter with a button click and view its output in the IDE console. The DLTK interpreter and console are closely entwined, and managing them in code is a complicated process. Before delving into all the classes and interfaces, here's a high-level look at the events that take place when the user clicks Run:

  1. The IDE responds to the button click by creating a launch object, collecting launch information, and adding a process to the launch.
  2. The creation of the launch notifies all launch listeners, including the DLTK console manager.
  3. The console manager acquires the launch information and uses it to create three objects: a console server, an interpreter object, and a console factory.
  4. The console server creates a socket (by default, port 25000) and waits for incoming connections.
  5. The console factory starts the interpreter and creates a console.
  6. The console manager accesses the Eclipse Console plug-in and adds the new console.

DLTK classes handle much of this processing, but you still have to create many of your own. The Octave IDE needs to support Octave-related launches, then it needs Octave-specific classes to manage the console and interpreter. This discussion presents each of these subjects in turn.

Configure the DLTK launch

In Eclipse, the process of starting an application is called launching the application in Run mode. Because Eclipse supports many launching executables, you must make sure that the Octave interpreter launches when the user executes an Octave script. This association is made through launch configuration types, and you can see the available types in the Run window by clicking Run > Run in Eclipse. Figure 11 shows what launch configuration types looks like.

Figure 11. The Octave launch configuration type
Screenshot shows the list of configuration thypes with Octave highlighted

To create a launch configuration type for Octave, the first step is to add an extension of the org.eclipse.debug.core.launchConfigurationTypes to plugin.xml. In Figure 11, you can see the Octave configuration type selected in the window. Listing 13 presents the extension that defines this new type.

Listing 13. Defining the Octave launch configuration type
<extension point="org.eclipse.debug.core.launchConfigurationTypes">

The example project executes Octave scripts but doesn't debug them. Therefore, the modes attribute is set to run but not run,debug. The most important attribute is delegate, which identifies a class that implements ILaunchConfigurationDelegate. This interface defines a single method, launch, which is called when the user creates a launch configuration and clicks Run.

DLTK provides its own implementation of ILaunchConfigurationDelegate called AbstractScriptLaunchConfigurationDelegate. When this class's launch method is invoked, it calls the methods listed in Table 9.

Table 9. The methods called during AbstractScriptLaunchConfigurationDelegate.launch()
createInterpreterConfigThis method constructs an InterpreterConfig object that receives all the information from the launch configuration.
getInterpreterRunnerThis method accesses the IInterpreterInstall object created by the interpreter installation type and uses it to form an IInterpreterRunner.
runRunnerThis method creates a process for the interpreter and adds it to the launch.

This process may sound complicated, but the first two methods simply organize information about the launch; the InterpreterConfig holds information taken from the launch configuration, and the IInterpreterInstall stores information taken from the preferences. It's important to understand that the third method does not actually start the interpreter. It creates an IProcess for the interpreter and adds it to the Launch object.

The DLTK console manager

When the DLTK plug-in org.eclipse.dltk.console.ui starts, it accesses the Eclipse Debug UI plug-in and adds an instance of ScriptConsoleManager as a launch listener. These listeners are notified whenever a launch is added, changed, or removed. The ScriptConsoleManager only responds to added launches, and its first task is to check whether the launch has a recognizable script nature. If so, the manager creates three important objects: a console server, an interpreter object, and a console factory.

The manager creates a ScriptConsoleServer to manage communication between the interpreter and the console. During its initialization, the server binds a ServerSocket to port 25000 and tells it to wait for incoming connection requests. These requests are embodied by ConsoleRequest objects, and as the server receives new requests, it calls the consoleConnected method of each.

After creating the server, the ScriptConsoleManager accesses a ConsoleRequest representing the interpreter. It does this by searching for an extension of the point org.eclipse.dltk.console.scriptInterpreter in plugin.xml. This extension point identifies a nature designation and a class that implements IScriptInterpreter, a subinterface of ConsoleRequest. After it obtains the server and interpreter object, the console manager can start creating the console. To do this, it calls upon a console factory.

The console factory and the console

Ordinarily, if you want to access an Eclipse console from a plug-in, you need only extend one extension point: org.eclipse.ui.console.consoleFactories. This extension must identify a class that implements IConsoleFactory. Then, when the user clicks Open Console in the Console view, Eclipse calls the class's openConsole method to create a new console.

DLTK makes things slightly more complicated. The DLTK ScriptConsoleManager also needs to access a console factory, but it requires a class that implements IScriptConsoleFactory. Conveniently, IConsoleFactory and IScriptConsoleFactory both require the same openConsole method. However, you need to create two extension points: one for the Eclipse console factory and one for the DLTK console factory. Listing 14 shows what these two points look like for the Octave IDE.

Listing 14. Identifying the Octave console factory
<extension point="org.eclipse.dltk.console.ui.scriptConsole">
      natureID="org.dworks.octaveide.nature" />
<extension point="org.eclipse.ui.console.consoleFactories">
      label="%OctaveConsole.Console" />

When the ScriptConsoleManager accesses a script console factory, it calls the factory's openConsole method, which creates the all-important ScriptConsole object. Like the Eclipse text editor, this console displays text using a document-based methodology; it has a ConsoleDocument, an IConsoleDocumentPartitioner, and arrays of Positions and Regions.

When the ScriptConsole is created, it calls setInterpreter with a reference to the IScriptInterpreter object that the console manager creates. This method creates a connection between the console and the interpreter in the form of an InitialStreamReader. This stream reader attaches itself to the interpreter by calling IScriptInterpreter.getInitialOutputStream. Next, the InitialStreamReader reads each line of the interpreter's output and sends it to the console's display. When it's done displaying text, the console prints a prompt and places itself in editable mode.

After creating the ScriptConsole, the ScriptConsoleManager accesses the Eclipse ConsoleManager and adds the new object to the list of available consoles. Now the new console will be visible to the user and accessible by the interpreter.

Run the interpreter

Let's take a closer look at the IScriptInterpreter object created by the ScriptConsoleManager. This interface extends the ConsoleRequest interface, which means that it can be delivered to the ScriptConsoleServer for processing. The console manager handles this delivery, and when the server receives the interpreter's request, it calls IScriptInterpreter.consoleConnected(IScriptConsoleIO protocol).

This method serves two important roles: It constructs the InitialStreamReader that enables data transfer between the console and the interpreter, and the method's argument identifies the protocol to be used for communication. The IScriptConsoleIO interface defines two important methods that receive data from the interpreter:

  • InterpreterResponse execInterpreter(String command). Sends a script command to the interpreter, then receives a response
  • ShellResponse execShell(String command, String[] args). Sends a shell command to the interpreter, then receives a response

These two functions direct commands to the interpreter and receive its output. Each IScriptConsoleIO protocol object is initialized with an InputStream and an OutputStream. When execInterpreter is called, the protocol sends the command to the interpreter to its OutputStream. When the interpreter responds, the protocol receives the InterpreterResponse (essentially, a String) and directs it to the console through the InputStream.

All communication to the interpreter relies on the protocol's methods. For example, the IScriptInterpreter interface contains the exec(String command) that sends commands to the interpreter. This command calls on IScriptConsoleIO.execInterpreter to deliver the command. Similarly, the interpreter's close method relies on IScriptConsoleIO.execShell.


As this tutorial has shown, the DLTK provides myriad features for creating development environments for dynamic languages. Creating these environments doesn't require a lot of code, but you have to know exactly what you're doing. To configure text presentation, you must understand how the editor, viewer, and document interact as well as the rudiments of rule-based partitioning and scanning. To enable user preferences, you have to grasp the strange class hierarchy of Eclipse preference pages and DLTK configuration blocks. To enable launching of dynamic-language projects, you have to deal with threads, streams, and the protocol methods needed to communicate between the console and script interpreter.

The DLTK learning curve is steep and continues upward for a significant distance. But if you need to build a full-featured IDE for your customized script language, you'll find no better toolset. In addition to the material presented in this tutorial, DLTK provides many sample IDEs for languages like PHP and Tcl.

Downloadable resources

Related topics


Sign in or register to add and subscribe to comments.

Zone=Open source
ArticleTitle=Build an Eclipse development environment for Perl, Python, and PHP